Selection of Solvents or Solvent Mixtures for Liquid–Liquid Extraction

Oct 13, 2014 - Selection of Solvents or Solvent Mixtures for Liquid–Liquid Extraction Using Predictive Thermodynamic Models or Access to the Dortmun...
2 downloads 12 Views 3MB Size
Article pubs.acs.org/IECR

Selection of Solvents or Solvent Mixtures for Liquid−Liquid Extraction Using Predictive Thermodynamic Models or Access to the Dortmund Data Bank Jürgen Gmehling*,†,‡ and Andre Schedemann†,‡ †

Technische Chemie, Universität Oldenburg, 26111 Oldenburg, Germany DDBST GmbH, Marie-Curie-Straße 10, 26129 Oldenburg, Germany



ABSTRACT: A sophisticated software package for the selection of the most suitable solvents or solvent mixtures for liquid− liquid extraction was developed. For the selection in particular, phase equilibrium information such as selectivity, capacity, etc. was used. For the determination of these properties, besides group contribution methods, such as UNIFAC or modified UNIFAC (Do) and the group contribution equations of state PSRK or VTPR, experimental data collected and updated in the Dortmund Data Bank (DDB) can be applied. For the final selection, pure component and mixture properties such as the surface tension, density, flash point, viscosity, etc. are accessible via DDB also. The applicability of the program package was demonstrated by means of various extraction processes of industrial importance.

1. INTRODUCTION A chemical process can in principle be structured into three steps, the preparation, reaction, and separation. Although the reactor can be considered the “heart” of a chemical plant, the major part of costs is usually caused by the separation step.1 In the preparation step and the separation step thermal separation processes, such as distillation, extraction, absorption, crystallization, membrane, and adsorption processes as well as mechanical processes are used to obtain the wanted products with the desired purity, to recycle unconverted reactants, and to remove the undesired side products and inert compounds. Because of the advantages (energy used for separation, large density differences between the liquid and vapor phase, and both phases being fluid), distillation processes are preferably applied for separation of mixtures in the chemical industry. This is even true for azeotropic systems, where the separation is realized by, for example, azeotropic or extractive distillation with the help of suitable solvents, pressure swing distillation, or hybrid processes. Also in crystallization processes, energy is used for separation. However, the handling of solid compounds can cause problems. In extraction and absorption processes, suitable solvents or solvent mixtures are applied for separation too.1 Extraction processes are of great importance if distillation cannot be applied successfully, for example, when the vapor pressure of the substances is too low. Extraction can also be applied when classes of compounds must be separated, for example, separation of aliphatics from aromatics, since for extraction processes, the vapor pressure has no influence on the separation. Extraction also shows advantages when dilute, highboiling components must be removed from a liquid carrier, for example, phenol from a wastewater stream or thermolabile components2 from fermentation broths. Extraction processes may also contribute to separation of azeotropic mixtures and mixtures with poor separation factors. When solvent-based separation processes are used, they sometimes compete with other solvent-based processes. In © XXXX American Chemical Society

separation of aliphatics from aromatics, for example, three different separation processes can be applied.1,2 When the composition of aromatics in the feed stream is beyond 90%, azeotropic distillation with acetone or methanol is the preferred separation technique. Extractive distillation is applied when the composition range is between 65% and 90% and when the feed composition has a narrow boiling temperature range. Liquid− liquid extraction is used when the concentration of aromatics is below 65%. Since the vapor pressure has no influence on the separation factor, extraction processes are applicable for separation of aliphatics from aromatics of C6−C12 hydrocarbon mixtures.1 For any solvent-based process, engineers must select the best suited solvent or solvent mixture. The software package of the Dortmund Data Bank (DDB)3 offers a tool for the selection of suitable solvents for azeotropic and extractive distillation,4,5 extraction,6 and absorption.7 This paper presents a new, extended software package8 that offers two different strategies for the selection of suitable solvents or solvent mixtures for separation by liquid−liquid extraction: (A) It supports selection of suitable solvents or solvent mixtures based on experimental phase equilibrium data stored in the DDB.3 (B) It supports search routines for suitable solvents and solvent mixtures with the help of prediction models.9 Use of experimental thermophysical properties stored in a factual database for the selection has the advantage that the results are very reliable; however, solvent selection is limited to the experimental data pool. The use of prediction models has the advantage that for the selection procedure, any solvent can be considered for which the required group interaction Received: July 21, 2014 Revised: October 7, 2014 Accepted: October 13, 2014

A

dx.doi.org/10.1021/ie502909k | Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Industrial & Engineering Chemistry Research

Article

denotes the extract stream, and R the raffinate stream. In this example, the solvent has a higher density and a lower volatility than the feed stream. An optimal solvent (extractant) must provide the best performance for the process. Criteria to select suitable solvents for extraction processes were suggested by Wauquier,13 Rydberg et al.,14 Perry et al.,15 Baerns et al.,1 and Cockrem et al.16 The most important properties such as miscibility gap, solvent loss, capacity, selectivity, etc. can be directly obtained from experimental data stored in a factual database or from predictive thermodynamic models. They can be summarized as follows: (A) Miscibility gap/solvent loss The existence of two liquid phases is a prerequisite for liquid−liquid extraction. The miscibility gap should be large, because this leads to a better separation performance and less solvent loss (see Figure 2). A large closed miscibility gap can handle feeds with a high amount of component A to be extracted (see Figure 2).15 The solvent loss, which depends on the size of the miscibility gap, can either directly be taken into account by the experimental solvent concentration in the raffinate phase or estimated with the activity coefficient of the solvent at infinite dilution in the raffinate phase17,18 (eq 1). A low solvent loss reduces the separation costs in the distillation step. 1 solvent loss ≈ ∞ γS,R (1)

parameters are available. By using predictive methods, an extended variety of solvents can be taken into account for selection, but of course the quality of the predicted separation factors, etc. is less accurate than the use of highly reliable experimental data. Group contribution methods show a few weaknesses,10 for example, isomer effects cannot be taken into account. Furthermore, poor results are obtained for the activity coefficients (solubilities) of alkanes, alkenes, and naphthenes with water, but the latter weakness can be avoided by a correction routine suggested by Banerjee.10,11 The use of both approaches is especially useful because they supplement each other. If no experimental data are available for a solvent for the given separation problem, a suitable solvent still can be found with the help of predictive group contribution methods. For solvents that cannot be specified with the group contribution method, experimental data may be available. Independent of the selected procedure for the final selection, several important properties (viscosity, surface tension, flash point, etc.) can additionally be determined with either experimental data or correlation parameters provided by the DDB. The search for suitable solvent mixtures extends the range of applicability of the software package. This option is important since capacity and selectivity are often antagonistic properties.12 Solvents with a high selectivity often show a low capacity and vice versa. When two solvents that show opposite behavior are mixed, improved properties may be obtained than for the pure solvents.

2. CRITERIA FOR SOLVENT SELECTION Extraction processes can be divided in different steps, the extraction step and the separation of the extract phase and the

(B) Capacity Another important property of solvents is the capacity. If the capacity is high, a lower solvent flow is required.15,19 A high capacity has also a direct positive influence on the selectivity (see eq 5). The capacity can be expressed by the solvent power Spi,20 which is equivalent to the inverse activity coefficient of the extracted component i at infinite dilution in the solvent S (see eq 2).21 The distribution coefficient Ki of the extracted component i, which is the ratio of the mole fraction of component i in the two liquid phases1 (see eq 3), can also be used as a measure for the capacity.19,22 1 Spi = ∞ γi,S (2)

Figure 1. General scheme of an extraction plant.1

Ki =

γi ,R γi ,E



xi ,E xi ,R

(3)

(C) Selectivity The solvent or solvent mixture should show a high selectivity to reduce the number of theoretical stages of the extraction step. The selectivity can either be estimated by the ratio Saij of the activity coefficients of the components to be separated (i and j) at infinite dilution in the solvent S20 (see eq 4) or by the 12,22 separation factor αext which is the ratio of the distribution ij , coefficients K of the components i and j (see eq 5). Figure 2. Two ternary miscibility gaps and their effect on the solvent loss and the possible feed compositions.

Saij ≈

raffinate phase. For a typical extraction plant, this is illustrated in Figure 1, where A is the component that must be extracted by the solvent S from a feed stream F consisting of A and B. E

αij ext = B

γi∞ ,S γj∞ ,S

(4)

Ki Kj

(5) dx.doi.org/10.1021/ie502909k | Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Industrial & Engineering Chemistry Research

Article

Figure 3. Start menu of the software package for the selection of solvents or solvent mixtures for liquid−liquid extraction (screenshot).

The interfacial tension can be estimated from the surface tensions σ. Small interfacial tensions lead to large mass transferarea, which accelerates the rate of extraction but also increases the risk of emulsion formation. A high density difference and low viscosities can counteract this risk.2 In the case of a high interfacial tension, the two liquid phases must be stirred to increase the interface,15 which is preferred over the formation of stable emulsions. (J) Flash point Extractants with high flash points are preferred for safety reasons. For the final selection further properties are important, but at present neither DDB nor predictive models do provide data such as (K) Toxicity The toxicity of the solvent should be low to decrease the risks for the staff of the plant. This information is available via the safety data sheet of the producer of the solvent. (L) Chemical stability The extractant should be inert against all components of the feed stream to avoid undesired reactions. Exceptions are desired reactions that support the extraction process, for example, the reaction of NaOH with phenol to sodium phenolate, etc. (M) Corrosivity The solvent should not be too corrosive. In this case, high investment costs caused by expensive materials for the construction of the plant (e.g., special steels, etc.) can be avoided. (N) Thermal stability The solvent should be stable at the selected extraction and distillation temperature. In this case, the solvent has not to be exchanged for a long time, and decomposition products do not appear. (O) Price The price of the solvent should be low. It is an important factor to account for the operation costs of the extraction step. (P) Foam formation

(D) Azeotropic behavior Azeotropes, at least homogeneous azeotropes, between the solvent (extractant) and the components to be separated should be avoided23 so that the solvent can be recovered easily by simple distillation from the extract and the raffinate stream. (E) Separation factors in the distillation columns The separation factors of the distillation steps αdis ij (eqs 6 and 7) give an idea about the cost of the distillation processes after the extraction step. Poor separation factors (α12 ≈ 1) should be avoided. dis

xS → 1

αi ,S

xS → 0

αi ,Sdis =

=

Ps γi∞ ,S i PSs

(6)

Pis ∞ s γS, i PS

(7)

(F) Density difference between the two liquid phases The two liquid phases can easily be separated if the density difference between the two liquid phases is sufficiently high. If the difference is too small but all other properties are satisfactory, centrifugal extractors must be used.23 If the densities of both liquid phases are nearly identical, separation cannot be realized.15 Other important properties for the final selection of the solvent or solvent mixture cannot be predicted with the required accuracy, but they can directly be taken from a factual database (e.g., DDB). (G) Melting point Solvents with a melting point Tm not far below the extraction temperature should not be used so that the precipitation of solids caused by a temperature decrease can be avoided. (H) Viscosity High dynamic viscosities η lead to a reduction of the mass transfer (diffusion coefficients).24 Therefore, high viscous solvents should not be chosen. (I) Surface tension C

dx.doi.org/10.1021/ie502909k | Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Industrial & Engineering Chemistry Research

Article

distillation columns, the separation factors are calculated at atmospheric pressure. After the determination of the different properties using either the selected predictive model or access to the Dortmund Data Bank, the solvents can be sorted by the value of the selection criterion (see eq 8). This consists of the distribution coefficient K (capacity) of the extracted components E(i) (ne components) and the separation factors αext E(i),R(j) between the extracted components E(i) and raffinate components R(i) (nr components). Furthermore, the solvent loss xS,R and the poorest value of αdis iS for the connected distillation columns as value for the “bottleneck” of the following separation by distillation can be taken into account (the reciprocal value of the separation factor, that is, αdis Si , is used for separation factors αdis iS smaller than one). Each of the contributions to the selection criterion can be weighted differently. The weighting factors, ws (selectivity), wc (capacity), wsl (solvent loss), and wsd (separation factor distillation), can be chosen by the user in the start menu. When, for example, a diluted high boiler should be extracted, following Cockrem et al.,16 the capacity is very important, since a high capacity reduces the solvent flow and reduces the costs of the separation of the extract stream by distillation.

Foam formation can be a problem, because it reduces the extraction speed in case of gas bubble inclusion. (Q) Water hazard class A high water hazard class leads to higher investment costs for the extraction plant to fulfill the specifications of environmental legislation.

3. SOFTWARE PACKAGE For the selection of the best suited solvents or solvent mixtures, the software package first uses the above-mentioned properties A−E. Exemplarily, the solvent selection for the separation of benzene, toluene, and n-heptane as model system for the separation of aromatics from aliphatics by extraction is used. Figure 3 shows the start menu of the software package. The user can define the components, which should be separated under the main topic “Components”. The DDB code numbers (e.g., 31 for benzene, 91 for n-heptane, and 161 for toluene) can be used directly. If the code numbers are unknown, the user can select the DDB code numbers with help of the “Comp.Sel.”-dialogue by using either the name, empirical formula, CAS number, melting point, normal boiling point, etc. When a component is diluted in the feed stream, it can be identified by a hook in the dialogue box on the right-hand side. In the “basic options”, the user can choose the extraction temperature. Furthermore, the user can specify the limits for, for example, the minimal separation factor of the extraction step and the maximum viscosity. With the dialogue “Solvent list”, a default list with common solvents,12,25−27 which are used for the selection, is opened. Of course, additional solvents can be added to this list. Also, all compounds of the DDB can be used as potential extractants (ca. 42 000). The user can also define binary solvent mixtures for the component selection in the “Mixture list”-dialogue. Then the user must specify the procedure (either a prediction model or access to DDB) for the selection of the solvent or solvent mixture. UNIFAC,28 modified UNIFAC (Do),29,30 and PSRK31 are models, which because of their comprehensive parameter matrices, can be used for a large number of solvents. These models have a large range of applicability, since the parameter matrices of these models are steadily revised and extended.32 The group contribution equation of state VTPR33,34 is an advanced predictive thermodynamic model, which can not only reliably predict the different phase equilibria and excess properties, but also other properties (enthalpies, densities, etc.) of the pure compounds and the mixtures. Main groups of a group contribution model can also be excluded, for example, when compounds with these functional groups are not stable or can react. In the example selected (see Figure 3), alkenes (2), aldehydes (10), and iodine compounds (32) are excluded. Of course, the experimental data stored in the DDB can directly be used to select suitable solvents for a given separation problem. For the selection, it provides about 80 000 activity coefficients at infinite dilution, approximately 28 000 LLE data sets and about 55 500 (a)zeotropic data points. Solvents that form a homogeneous azeotrope or show a low flash point can be excluded with “Additional Options”. The user can also decide if solvent mixtures should be considered. If so, the software package is able to look for suitable solvent mixtures on the basis of the results for the pure solvents. The distribution coefficients can also be calculated considering the molar mass,35,36 which prefers solvents with a low molar mass during the selection. To account for the costs in the connected

⎡ ne nr K ⎤ws ne ⎢∑ ∑ E(i) /(nc − 1)⎥ ·[∑ K /ne]wc ·(1 − x )wsl ·log(α dis)wsd E(i) S ,R iS ⎢⎣ ⎥⎦ KR(j) i=1 i=1 j=1 = Selection Criterion

(8)

4. SOLVENT SELECTION BASED ON PREDICTION MODELS First, the procedure using prediction models should be described exemplarily for the separation of the system

Figure 4. Important points of a ternary liquid−liquid equilibrium used for solvent selection.

benzene/n-heptane/toluene by extraction. The program starts to collect the pure component properties of the components to be separated. Then, for later calculations, the activity coefficients and activity coefficients at infinite dilution of the different compounds in the feed are estimated with the help of the chosen prediction model. These calculations should be explained exemplarily for a feed containing benzene, n-heptane, and small amounts of toluene. While the activity coefficient at infinite dilution for toluene is calculated for an equimolar mixture of benzene and n-heptane, the activity coefficients for benzene (n-heptane) are calculated at infinite dilution in nheptane (benzene). In the second example, the feed is composed of small amounts of toluene and benzene in nheptane. In this case, the activity coefficients of toluene and D

dx.doi.org/10.1021/ie502909k | Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Industrial & Engineering Chemistry Research

Article

Table 1. First Part of the Result Sheet of the Solvent Selection for the Benzene (31)/n-Heptane (91)/Toluene (161) Separation Containing the Main Information about the Selected Solvents (Predicted Using Modified UNIFAC (Do) at 300 K; Default Values are Given in Figure 3) solvent (S)

distribution

distribution coefficient

separation factor

ranking no.

DDB no.

name

(1)

(2)

(3)

K(1)

K(2)

K(3)

K(1)/K(2)

K(1)/K(3)

sel. crit.

1 2 3 4 5 6 7 8 9 19 20 24 29

1285 1160 443 1711 151 284 563 24 72 3 227 174 614

tetraethylene glycol 2-pyrrolidone triethylene glycol phthalic acid dimethyl ester dimethyl sulfoxide N-methyl-2-pyrrolidone ethylene glycol diacetate benzyl alcohol N,N-dimethylformamide acetonitrile N,N-dimethylacetamide water 1,4-butanediol

31 31 31 161 31 31 31 31 31 31 31 31 31

161/ 161/ 161/ 31 / 161/ 161/ 161/ 161/ 161/ 161/ 161/ 161/ 161/

91 91 91 91 91 91 91 91 91 91 91 91 91

0.8716 0.6106 0.4957 1.4407 0.4843 1.5423 1.2885 0.8922 0.9573 0.572 1.3178 0.0007 0.2246

0.7062 0.4178 0.3944 1.2678 0.3625 1.3123 1.0134 0.8014 0.7774 0.4517 1.1322 0.0002 0.1827

0.064 0.0246 0.0279 0.2828 0.0106 0.1535 0.1808 0.1524 0.089 0.0522 0.3088 →0 0.023

1.2 1.5 1.3 1.1 1.3 1.2 1.3 1.1 1.2 1.3 1.2 3.5 1.2

14.0 25.0 18.0 5.1 46.0 10.0 7.1 5.9 11.0 11.0 4.3 1100 9.8

14.0 9.1 7.7 7.5 7.2 6.5 4.1 3.0 2.2 0.9 0.8 0.38 0.2

γi∞ ,F γi∞ ,S

αij ext,pre =



K i pre Kj pre

xi ,S xi ,F

low DD

x x

x

The final distribution coefficients and the separation factors can be determind with the help of these calculations (see eqs 3 and 5). Point E represents the solvent-rich extract phase, and R represents the raffinate phase. When more than two components should be separated, the program determines the optimal component distribution in the extract and raffinate phase with the help of the highest average separation factor αext ij . According to Hampe,12 the calculation of the separation factor at infinite dilution of the extracted components shows advantages in comparison to calculations at higher concentrations. Normally, the highest values for the separation factors are obtained at infinite dilution. Furthermore, most extraction steps are required at low composition of the extracted component. The calculation at infinite dilution simplifies a comparison of the results. For the following ranking, the selection criterion (eq 8) is determined for each solvent. The calculations of this criterion are given below for the benzene/n-heptane/toluene separation with dimethyl sulfoxide (DMSO) and N-methyl-2-pyrrolidone (NMP) with the weighting factors selected in the start menu (ws = 1; wc = 1; wsl = 3; wsd = 0.75) (values from Table 1−3; solvent loss (xS,R) is represented by 1 − xRaffinate(3)):

benzene are calculated at infinite dilution in n-heptane, while the activity coefficient of n-heptane is set to one. For the solvent selection, each of the solvents is considered independently. First, the program checks the pure component properties (viscosity, surface tension, etc.) of the pure solvent. When limits, for example, maximum viscosity, melting temperature, flash point, etc., are not fulfilled, the solvent is removed from the list of potential extractants, and the next solvent is considered. Otherwise, the program calculates the activity coefficients of each component at infinite dilution in the solvent S with the chosen prediction model. With the help of these activity coefficients and the already calculated activity coefficients in the feed F, preliminary values of Ki and αext ij can be calculated (see eqs 9 and 10), and the components are sorted by the values of their distribution coefficients. The calculation of only provisional values18 reduces the needed computer time for this step. K i pre =

diff. sep.

(9)

(10)

When at least one separation factor is higher than the default value, the program starts to check for binary liquid−liquid equilibrium (LLE) between all feed components and the solvent. In the same routine, the program checks if azeotropic behavior occurs and calculates the separation factor of the vapor−liquid equilibrium for these binary systems. When LLE behavior for one of the binary systems was found, the program searches for the key component in the raffinate phase (component B). It is the component that shows the lowest distribution coefficient, is not diluted, and shows a miscibility gap with the solvent. The activity coefficients of all components can then be calculated in the two liquid phases of the binary LLE between the solvent and the key component (component B) (points E and R in Figure 4). The activity coefficients of the other components (e.g., component A in Figure 4), which are not the solvent or the key component, are calculated at infinite dilution in the two liquid phases.15

⎡ 0.4863 + 0.3625 ⎤1 1 0.0106 ⎥ ⎡ 0.4843 + 0.3625 ⎤ ⎢ 0.0106 · · DMSO: ⎢ ⎢ ⎥ ⎥ ⎣ ⎦ 2 2 ⎣ ⎦ (1 − 0.011)3 · log(1/0.46)0.75 = 7.2

(

)

1 1.3123 ⎤ ⎡ 1.54233 1 + ⎢ 0.1535 0.1535 ⎥ ⎡ 1.54233 + 1.3123 ⎤ ·⎢ · NMP: ⎢ ⎥ ⎥ ⎦ ⎣ 2 2 ⎢⎣ ⎥⎦ (1 − 0.112)3 · log(4.1)0.75 = 6.5

(

)

The results show that DMSO is ranked higher using the given weighting factors. When the capacity is weighted stronger, the order can change (e.g., ws = 1; wc = 1.5; wsl = 3; wsd = 0.75): E

dx.doi.org/10.1021/ie502909k | Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

320 7.2 13 87 0.46 4.1 4 4 0.52 0.06 1.7 0.059 0.88 1200 22 37 1900 4 13 12 11 1.6 0.15 3.1 0.022 5 45 48 58 81 14 9.8 6.8 11 5.8 1.3 5.1 1600 29 9600 120 150 420 12 62 87 39 6.3 0.41 10 0.055 12 55 70 72 46 21 17 7.3 17 10 2.7 8.6 1100 42

150 230 230 37 63 40 23 12 23 7 22 5700 41

α(3)/(S) xS → 0 α(3)/(S) xS → 1 α(2)/(S) xS → 0

xExtract(3)

xRaffinate(3)

extract phase

raffinate phase

1 2 3 4 5 6 7 8 9 19 20 24 29

0.064 0.024 0.028 0.265 0.01 0.136 0.17 0.149 0.084 0.049 0.25 0 0.023

0.998 0.984 0.998 0.936 0.989 0.888 0.941 0.975 0.941 0.945 0.809 0.991 1

1.098 1.087 1.105 1.063 1.086 0.958 1.022 0.969 0.904 0.763 0.847 0.994 1.001

0.681 0.684 0.681 0.716 0.682 0.706 0.702 0.686 0.688 0.682 0.713 0.68 0.68

⎡ 0.4863 + 0.3625 ⎤1 1.5 0.0106 ⎥ ⎡ 0.4843 + 0.3625 ⎤ ⎢ 0.0106 · DMSO: ⎢ ⎥⎦ · ⎥ ⎢⎣ 2 2 ⎣ ⎦ (1 − 0.011)3 · log(1/0.46)0.75 = 4.7

het het

het

)

At the end, all solvents are sorted with the help of the selection criterion as shown in Tables 1−4. The output should be explained for the benzene/n-heptane/toluene example. Because of space reasons, only the results for a few of the suitable solvents are listed. In the first column of Table 1, the ranking place of the solvents is given. In first place, the solvent with the highest value of the selection criteria can be found, that is, the solvent that shows the best performance for the given separation problem and weighting factors. The other solvents follow with decreasing value of the selection criterion. Then, the names and the DDB numbers of the pure solvents are listed. In the following column, the components are shown that are extracted (left-hand side of the slash) and that stay in the raffinate phase (right-hand side of the slash). Their numbering from 1−3 from the table header is used later for the other values, for example, separation factors. Furthermore, the distribution coefficients of the components are given. The next column lists the separation factor for the extraction of each component on the basis of the best soluble component. In the last columns, the values for the selection criterion are given. Furthermore, the program provides information about separation problems in the connected distillation columns (diff. sep) or caused by a very low density difference between the two liquid phases (low DD). The exact information about these problems is given in the next tables. It is necessary to mention that the distribution coefficient of n-heptane in water (solvent number 24) has been corrected by the Banerjee approach10,11 so that the prediction provides realistic values for the distribution coefficients and separation factors. In the first places are solvents that show a good overall performance (for

none none none none none none none none none 0.467 none 0.3 het none

none none none none none none none none none 0.889 none 0.561 het none

none none none none 0.046 none none none 0.116 0.629 none 0.464 0.004

het

)

1 1.3123 ⎤ ⎡ 1.54233 1.5 ⎢ 0.1535 + 0.1535 ⎥ ⎡ 1.54233 + 1.3123 ⎤ · · NMP: ⎢ ⎥ ⎢ ⎥ ⎣ ⎦ 2 2 ⎢⎣ ⎥⎦ (1 − 0.112)3 · log(4.1)0.75 = 7.7